Predicting when Divvy bike share stations will be empty or full

Fellows: Walter Dempsey, Adam Fishman, Jette Henderson, Breanna Miller, Vidhur Vohra
Data Science Mentor(s): Juan-Pablo Velez
Project Partner: Chicago Department of Transportation
[Github Repository]

In 2013, the City of Chicago launched Divvy, a new bike share system designed to connect people to transit, and to make short one-way trips across town easy. Bike share is citywide bike rental – you can take a bike out at a station on one street corner and drop it off at another.

Popular in Europe and Asia, bike share has landed in the United States: Boston, DC, San Francisco and New York launched systems in the past few years.

These systems share a central flaw: because of commuting patterns, bikes tend to pile up downtown in morning and on the outskirts in the afternoon. This imbalance can make using bikeshare difficult, because people can’t take out bikes from empty stations, or finish their rides at full stations.

To prevent this problem, bikeshare operators drive trucks around to reallocate bikes from full stations to empty ones. But they can only see the current number of bikes at each station – not how many will be there in an hour or two.

We worked with the City of Chicago’s Department of Transportation to change this by analyzing weather and bikeshare station trends to predict how many bikes are likely to be at each Divvy station in the future.

By using predictive analytics, Divvy staff is now able to rebalance bikes proactively across the system, ensuring there’s always a bike there when you need it – and making bike share a first-class mode of transportation in Chicago and beyond.